2. Restraint system components and working principles

An occupant restraint system is not a collection of independent safety devices; it is a coordinated energy-management chain. Every component — belt webbing, retractor, pretensioner, load limiter, airbag module, and child restraint — is tuned to act within a specific time window and force range so that the occupant decelerates with the vehicle structure rather than against it. In CAE correlation work, understanding this chain is what allows an engineer to decide whether a mismatch between simulation and test originates in the belt model, the airbag inflator, or the occupant positioning.

Seat belts: the primary load path

The three-point belt remains the dominant restraint because it couples the upper torso and pelvis into the seat and vehicle structure. Its behavior is governed by three interacting elements: the retractor, which locks on webbing acceleration or vehicle deceleration (typically sensed above roughly 0.45 g spool acceleration or 1.5 g vehicle deceleration in production designs); the D-ring and buckle geometry, which set the belt path and therefore the angle at which load enters the occupant; and the webbing, whose elongation under load contributes compliance. In a model, the belt is usually represented as a membrane or 1D seatbelt element set with a defined force–strain curve, routed through slip rings at the D-ring and buckle latch. The critical modeling detail is friction and slip: a belt that slips too freely at the D-ring will under-predict shoulder load and over-predict chest deflection.

Belt fit is a first-order variable. A belt routed over the lateral edge of the shoulder rather than the clavicle produces a different load vector and is a common source of correlation error when the dummy or human body model is seated slightly out of position relative to the test setup.

Pretensioners: removing slack before the crash pulse peaks

A pyrotechnic or motorized pretensioner retracts webbing in the first 10–30 ms after triggering, removing slack and coupling the occupant to the restraint earlier. Buckle pretensioners pull the latch down and inboard; retractor pretensioners spool webbing directly. The effect is a steeper, earlier rise in belt force — visible in the force–time trace as a shortened flat "slack" region before the main loading ramp. In simulation, pretensioner action is applied as a prescribed retraction displacement or force over a defined time window. If the pretensioner is modeled but the retraction stroke is too small, the model will show late occupant engagement and under-predicted forward excursion.

Load limiters: capping chest load

Once the pretensioner has coupled the occupant, the load limiter controls how much force the belt is allowed to transmit to the chest. A torsion-bar retractor or a tearing stitch in the webbing permits controlled webbing payout above a threshold force, typically in the range of 3–6 kN at the shoulder for a mid-size adult occupant. The design trade-off is direct: a higher limit reduces forward excursion and injury risk to the head and lower extremities, while a lower limit reduces chest deflection and rib acceleration. In CAE, the limiter is represented by a force-limiting element in the belt path or by a retractor payout curve. Correlation failures here usually appear as a plateau in the shoulder force trace at the wrong level or with the wrong onset time.

Airbags: supplemental, not primary

Frontal airbags are calibrated to deploy into the gap between the occupant and the interior after the belt has begun loading the occupant. Inflator output, vent area, and module geometry determine the bag's pressure–time history and therefore the force it applies to the head and chest. A typical driver airbag reaches full inflation in roughly 25–35 ms and begins venting immediately. The bag is a supplemental restraint: it is tuned to cushion the head and distribute chest load, not to stop the occupant alone. In models, the airbag is a control-volume or corpuscular-particle (gas-flow) model coupled to the occupant via contact. The most sensitive parameters are vent coefficient, inflator mass-flow curve, and bag-to-occupant gap at trigger time.

Out-of-position (OOP) scenarios are where airbag modeling earns its keep. A bag deploying against a closely positioned occupant produces high local pressure and can generate injurious loads; this is why suppression and low-risk deployment logic exists, and why OOP test matrices are part of correlation campaigns.

Child restraint systems

Child restraints change the load path fundamentally. A rear-facing infant seat distributes crash load across the shell and harness rather than through a three-point belt on the child; a forward-facing seat with a five-point harness couples the child through the harness and top tether. ISOFIX/LATCH anchorages and top tethers limit seat rotation. In CAE, CRS models require accurate shell stiffness, harness routing and friction, and anchor compliance. The occupant is a child dummy (e.g., Q-series), whose biofidelity and positioning differ substantially from adult ATDs — a frequent source of correlation drift when the same modeling assumptions are carried over from adult cases.

How the components interact

The timing relationship is the heart of the system. Pretensioner fires first, removing slack; the belt then loads the occupant while the load limiter caps chest force; the airbag deploys into the developing gap and cushions the head as forward excursion peaks; the vehicle structure absorbs the remaining energy. If any element is early, late, or mis-scaled, the occupant kinematics shift and the correlation degrades. The diagram below shows this coordinated sequence and the observable signal each stage produces.

Practical correlation checkpoints

When a sled or full-vehicle test is compared against the model, the component chain gives a structured diagnostic order. First, align the trigger time: a 2 ms shift in pretensioner or airbag fire time moves every downstream signal. Second, compare the shoulder and lap belt force traces shape by shape — slack region, ramp slope, plateau level, and unloading. Third, compare head acceleration and chest deflection against the airbag pressure trace. A model that matches belt force but over-predicts head acceleration usually has an airbag vent or inflator issue; a model that matches head acceleration but over-predicts chest deflection usually has a limiter threshold set too high.

For child restraint cases, add a fourth check: harness routing and anchor compliance. A CRS model that under-predicts seat rotation typically has a top tether modeled as rigid, or an ISOFIX anchor without the correct compliance. These are not cosmetic differences — they change the child's excursion and the head impact location.

The practical takeaway is that restraint correlation is a chain-matching exercise. Each component has a characteristic signature in the measured signals, and the engineer's job is to read those signatures in the correct order — trigger timing, belt force, limiter plateau, airbag pressure, occupant response — so that adjustments are made at the right link rather than at the end of the chain.